annotations4all is a modular framework for performing text annotation tasks with Large Language Models (LLMs) in a reproducible workflow. While recent studies have shown that contextual prompting allows LLMs to achieve high-quality Named Entity Recognition (NER) without fine-tuning, most implementations remain ad-hoc and difficult to reproduce. annotations4all provides a configurable pipeline that separates prompt templates, model interaction, response parsing, and evaluation. It supports both local and API-based models and produces structured span annotations that can be exported in standard formats. The framework was developed in the context of the NER4All project, where contextual prompting outperformed established NER tools such as spaCy and flair on historical texts. annotations4all generalizes these methods into reusable components and enables researchers to apply LLM-based annotation workflows to new corpora and domains.